Downloads · 30 days
21
57% of all-time downloads
theoriclabs/qwen3-0.6b-pokemon-showdown
qwen3-0.6b-pokemon-showdown is a machine learning model from theoriclabs. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.
A LoRA adapter that turns Qwen3-0.6B into a gen9randombattle player. The battle state is text with numbered options; the model answers one digit and the player takes the highest-probability legal digit. Trained by imi…
Downloads · 30 days
21
57% of all-time downloads
All-time downloads
37
Public
Repo size
51.8 MB
Likes
1
Public
Click a slice to open those files.
.safetensors40.4 MB · 72%
From the Hugging Face model README
A LoRA adapter that turns Qwen3-0.6B into a gen9randombattle player. The battle state is
text with numbered options; the model answers one digit and the player takes
the highest-probability legal digit. Trained by imitation on
SimpleHeuristicsPlayer decisions (poke-env), tera off.
Win rate against poke-env's MaxBasePowerPlayer, 200 battles each:
{
"random": {
"won": 13,
"battles": 200,
"win_rate": 0.065,
"win_rate_95ci": [
0.0384,
0.108
]
},
"expert_no_tera": {
"won": 176,
"battles": 200,
"win_rate": 0.88,
"win_rate_95ci": [
0.8277,
0.918
]
},
"qwen_untuned": {
"won": 23,
"battles": 200,
"win_rate": 0.115,
"win_rate_95ci": [
0.0779,
0.1666
]
},
"qwen_sft": {
"won": 179,
"battles": 200,
"win_rate": 0.895,
"win_rate_95ci": [
0.8448,
0.9303
]
}
}
The prompt builder, player, and training loop are in train.py. Guide:
https://letsusecompute.com/posts/qwen-pokemon